| Message ID | e2682f84-b3bb-b9fb-edd8-863b9036de95@suse.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | IOREQ: mapcache invalidation request sending corrections | expand |
> -----Original Message----- > From: Jan Beulich <jbeulich@suse.com> > Sent: 02 February 2021 15:15 > To: xen-devel@lists.xenproject.org > Cc: Andrew Cooper <andrew.cooper3@citrix.com>; Wei Liu <wl@xen.org>; Roger Pau Monné > <roger.pau@citrix.com>; Paul Durrant <paul@xen.org>; Julien Grall <julien@xen.org>; Stefano Stabellini > <sstabellini@kernel.org>; George Dunlap <george.dunlap@citrix.com> > Subject: [PATCH v2 2/2] IOREQ: refine when to send mapcache invalidation request > > XENMEM_decrease_reservation isn't the only means by which pages can get > removed from a guest, yet all removals ought to be signaled to qemu. Put > setting of the flag into the central p2m_remove_page() underlying all > respective hypercalls as well as a few similar places, mainly in PoD > code. > > Additionally there's no point sending the request for the local domain > when the domain acted upon is a different one. The latter domain's ioreq > server mapcaches need invalidating. We assume that domain to be paused > at the point the operation takes place, so sending the request in this > case happens from the hvm_do_resume() path, which as one of its first > steps calls handle_hvm_io_completion(). > > Even without the remote operation aspect a single domain-wide flag > doesn't do: Guests may e.g. decrease-reservation on multiple vCPU-s in > parallel. Each of them needs to issue an invalidation request in due > course, in particular because exiting to guest context should not happen > before the request was actually seen by (all) the emulator(s). > > Signed-off-by: Jan Beulich <jbeulich@suse.com> > --- > v2: Preemption related adjustment split off. Make flag per-vCPU. More > places to set the flag. Also handle acting on a remote domain. > Re-base. I'm wondering if a per-vcpu flag is overkill actually. We just need to make sure that we don't miss sending an invalidation where multiple vcpus are in play. The mapcache in the emulator is global so issuing an invalidate for every vcpu is going to cause an unnecessary storm of ioreqs, isn't it? Could we get away with the per-domain atomic counter? Paul
On 04.02.2021 10:26, Paul Durrant wrote: > > >> -----Original Message----- >> From: Jan Beulich <jbeulich@suse.com> >> Sent: 02 February 2021 15:15 >> To: xen-devel@lists.xenproject.org >> Cc: Andrew Cooper <andrew.cooper3@citrix.com>; Wei Liu <wl@xen.org>; Roger Pau Monné >> <roger.pau@citrix.com>; Paul Durrant <paul@xen.org>; Julien Grall <julien@xen.org>; Stefano Stabellini >> <sstabellini@kernel.org>; George Dunlap <george.dunlap@citrix.com> >> Subject: [PATCH v2 2/2] IOREQ: refine when to send mapcache invalidation request >> >> XENMEM_decrease_reservation isn't the only means by which pages can get >> removed from a guest, yet all removals ought to be signaled to qemu. Put >> setting of the flag into the central p2m_remove_page() underlying all >> respective hypercalls as well as a few similar places, mainly in PoD >> code. >> >> Additionally there's no point sending the request for the local domain >> when the domain acted upon is a different one. The latter domain's ioreq >> server mapcaches need invalidating. We assume that domain to be paused >> at the point the operation takes place, so sending the request in this >> case happens from the hvm_do_resume() path, which as one of its first >> steps calls handle_hvm_io_completion(). >> >> Even without the remote operation aspect a single domain-wide flag >> doesn't do: Guests may e.g. decrease-reservation on multiple vCPU-s in >> parallel. Each of them needs to issue an invalidation request in due >> course, in particular because exiting to guest context should not happen >> before the request was actually seen by (all) the emulator(s). >> >> Signed-off-by: Jan Beulich <jbeulich@suse.com> >> --- >> v2: Preemption related adjustment split off. Make flag per-vCPU. More >> places to set the flag. Also handle acting on a remote domain. >> Re-base. > > I'm wondering if a per-vcpu flag is overkill actually. We just need > to make sure that we don't miss sending an invalidation where > multiple vcpus are in play. The mapcache in the emulator is global > so issuing an invalidate for every vcpu is going to cause an > unnecessary storm of ioreqs, isn't it? The only time a truly unnecessary storm would occur is when for an already running guest mapcache invalidation gets triggered by a remote domain. This should be a pretty rare event, so I think the storm in this case ought to be tolerable. > Could we get away with the per-domain atomic counter? Possible, but quite involved afaict: The potential storm above is the price to pay for the simplicity of the model. It is important to note that while we don't need all of the vCPU-s to send these ioreqs, we need all of them to wait for the request(s) to be acked. And this waiting is what we get "for free" using the approach here, whereas we'd need to introduce new logic for this with an atomic counter (afaict at least). Jan
Paul (or others), thoughts? On 04.02.2021 12:24, Jan Beulich wrote: > On 04.02.2021 10:26, Paul Durrant wrote: >>> From: Jan Beulich <jbeulich@suse.com> >>> Sent: 02 February 2021 15:15 >>> >>> XENMEM_decrease_reservation isn't the only means by which pages can get >>> removed from a guest, yet all removals ought to be signaled to qemu. Put >>> setting of the flag into the central p2m_remove_page() underlying all >>> respective hypercalls as well as a few similar places, mainly in PoD >>> code. >>> >>> Additionally there's no point sending the request for the local domain >>> when the domain acted upon is a different one. The latter domain's ioreq >>> server mapcaches need invalidating. We assume that domain to be paused >>> at the point the operation takes place, so sending the request in this >>> case happens from the hvm_do_resume() path, which as one of its first >>> steps calls handle_hvm_io_completion(). >>> >>> Even without the remote operation aspect a single domain-wide flag >>> doesn't do: Guests may e.g. decrease-reservation on multiple vCPU-s in >>> parallel. Each of them needs to issue an invalidation request in due >>> course, in particular because exiting to guest context should not happen >>> before the request was actually seen by (all) the emulator(s). >>> >>> Signed-off-by: Jan Beulich <jbeulich@suse.com> >>> --- >>> v2: Preemption related adjustment split off. Make flag per-vCPU. More >>> places to set the flag. Also handle acting on a remote domain. >>> Re-base. >> >> I'm wondering if a per-vcpu flag is overkill actually. We just need >> to make sure that we don't miss sending an invalidation where >> multiple vcpus are in play. The mapcache in the emulator is global >> so issuing an invalidate for every vcpu is going to cause an >> unnecessary storm of ioreqs, isn't it? > > The only time a truly unnecessary storm would occur is when for > an already running guest mapcache invalidation gets triggered > by a remote domain. This should be a pretty rare event, so I > think the storm in this case ought to be tolerable. > >> Could we get away with the per-domain atomic counter? > > Possible, but quite involved afaict: The potential storm above > is the price to pay for the simplicity of the model. It is > important to note that while we don't need all of the vCPU-s > to send these ioreqs, we need all of them to wait for the > request(s) to be acked. And this waiting is what we get "for > free" using the approach here, whereas we'd need to introduce > new logic for this with an atomic counter (afaict at least). > > Jan >
On 17/02/2021 08:30, Jan Beulich wrote: > Paul (or others), thoughts? > > On 04.02.2021 12:24, Jan Beulich wrote: >> On 04.02.2021 10:26, Paul Durrant wrote: >>>> From: Jan Beulich <jbeulich@suse.com> >>>> Sent: 02 February 2021 15:15 >>>> >>>> XENMEM_decrease_reservation isn't the only means by which pages can get >>>> removed from a guest, yet all removals ought to be signaled to qemu. Put >>>> setting of the flag into the central p2m_remove_page() underlying all >>>> respective hypercalls as well as a few similar places, mainly in PoD >>>> code. >>>> >>>> Additionally there's no point sending the request for the local domain >>>> when the domain acted upon is a different one. The latter domain's ioreq >>>> server mapcaches need invalidating. We assume that domain to be paused >>>> at the point the operation takes place, so sending the request in this >>>> case happens from the hvm_do_resume() path, which as one of its first >>>> steps calls handle_hvm_io_completion(). >>>> >>>> Even without the remote operation aspect a single domain-wide flag >>>> doesn't do: Guests may e.g. decrease-reservation on multiple vCPU-s in >>>> parallel. Each of them needs to issue an invalidation request in due >>>> course, in particular because exiting to guest context should not happen >>>> before the request was actually seen by (all) the emulator(s). >>>> >>>> Signed-off-by: Jan Beulich <jbeulich@suse.com> >>>> --- >>>> v2: Preemption related adjustment split off. Make flag per-vCPU. More >>>> places to set the flag. Also handle acting on a remote domain. >>>> Re-base. >>> >>> I'm wondering if a per-vcpu flag is overkill actually. We just need >>> to make sure that we don't miss sending an invalidation where >>> multiple vcpus are in play. The mapcache in the emulator is global >>> so issuing an invalidate for every vcpu is going to cause an >>> unnecessary storm of ioreqs, isn't it? >> >> The only time a truly unnecessary storm would occur is when for >> an already running guest mapcache invalidation gets triggered >> by a remote domain. This should be a pretty rare event, so I >> think the storm in this case ought to be tolerable. >> >>> Could we get away with the per-domain atomic counter? >> >> Possible, but quite involved afaict: The potential storm above >> is the price to pay for the simplicity of the model. It is >> important to note that while we don't need all of the vCPU-s >> to send these ioreqs, we need all of them to wait for the >> request(s) to be acked. And this waiting is what we get "for >> free" using the approach here, whereas we'd need to introduce >> new logic for this with an atomic counter (afaict at least). >> >> Jan >> > Ok, let's take the patch as-is then. Reviewed-by: Paul Durrant <paul@xen.org>
On 02.02.2021 16:14, Jan Beulich wrote: > XENMEM_decrease_reservation isn't the only means by which pages can get > removed from a guest, yet all removals ought to be signaled to qemu. Put > setting of the flag into the central p2m_remove_page() underlying all > respective hypercalls as well as a few similar places, mainly in PoD > code. > > Additionally there's no point sending the request for the local domain > when the domain acted upon is a different one. The latter domain's ioreq > server mapcaches need invalidating. We assume that domain to be paused > at the point the operation takes place, so sending the request in this > case happens from the hvm_do_resume() path, which as one of its first > steps calls handle_hvm_io_completion(). > > Even without the remote operation aspect a single domain-wide flag > doesn't do: Guests may e.g. decrease-reservation on multiple vCPU-s in > parallel. Each of them needs to issue an invalidation request in due > course, in particular because exiting to guest context should not happen > before the request was actually seen by (all) the emulator(s). > > Signed-off-by: Jan Beulich <jbeulich@suse.com> > --- > v2: Preemption related adjustment split off. Make flag per-vCPU. More > places to set the flag. Also handle acting on a remote domain. > Re-base. Can I get an Arm side ack (or otherwise) please? Thanks, Jan > --- a/xen/arch/arm/p2m.c > +++ b/xen/arch/arm/p2m.c > @@ -759,10 +759,9 @@ static void p2m_free_entry(struct p2m_do > * has failed (error case). > * So, at worst, the spurious mapcache invalidation might be sent. > */ > - if ( (p2m->domain == current->domain) && > - domain_has_ioreq_server(p2m->domain) && > - p2m_is_ram(entry.p2m.type) ) > - p2m->domain->mapcache_invalidate = true; > + if ( p2m_is_ram(entry.p2m.type) && > + domain_has_ioreq_server(p2m->domain) ) > + ioreq_request_mapcache_invalidate(p2m->domain); > #endif > > p2m->stats.mappings[level]--; > --- a/xen/arch/arm/traps.c > +++ b/xen/arch/arm/traps.c > @@ -1509,8 +1509,8 @@ static void do_trap_hypercall(struct cpu > * Note that sending the invalidation request causes the vCPU to block > * until all the IOREQ servers have acknowledged the invalidation. > */ > - if ( unlikely(curr->domain->mapcache_invalidate) && > - test_and_clear_bool(curr->domain->mapcache_invalidate) ) > + if ( unlikely(curr->mapcache_invalidate) && > + test_and_clear_bool(curr->mapcache_invalidate) ) > ioreq_signal_mapcache_invalidate(); > #endif > } > --- a/xen/arch/x86/hvm/hypercall.c > +++ b/xen/arch/x86/hvm/hypercall.c > @@ -32,7 +32,6 @@ > > static long hvm_memory_op(int cmd, XEN_GUEST_HANDLE_PARAM(void) arg) > { > - const struct vcpu *curr = current; > long rc; > > switch ( cmd & MEMOP_CMD_MASK ) > @@ -42,14 +41,11 @@ static long hvm_memory_op(int cmd, XEN_G > return -ENOSYS; > } > > - if ( !curr->hcall_compat ) > + if ( !current->hcall_compat ) > rc = do_memory_op(cmd, arg); > else > rc = compat_memory_op(cmd, arg); > > - if ( (cmd & MEMOP_CMD_MASK) == XENMEM_decrease_reservation ) > - curr->domain->mapcache_invalidate = true; > - > return rc; > } > > @@ -327,9 +323,11 @@ int hvm_hypercall(struct cpu_user_regs * > > HVM_DBG_LOG(DBG_LEVEL_HCALL, "hcall%lu -> %lx", eax, regs->rax); > > - if ( unlikely(currd->mapcache_invalidate) && > - test_and_clear_bool(currd->mapcache_invalidate) ) > + if ( unlikely(curr->mapcache_invalidate) ) > + { > + curr->mapcache_invalidate = false; > ioreq_signal_mapcache_invalidate(); > + } > > return curr->hcall_preempted ? HVM_HCALL_preempted : HVM_HCALL_completed; > } > --- a/xen/arch/x86/mm/p2m.c > +++ b/xen/arch/x86/mm/p2m.c > @@ -28,6 +28,7 @@ > #include <xen/vm_event.h> > #include <xen/event.h> > #include <xen/grant_table.h> > +#include <xen/ioreq.h> > #include <xen/param.h> > #include <public/vm_event.h> > #include <asm/domain.h> > @@ -815,6 +816,8 @@ p2m_remove_page(struct p2m_domain *p2m, > } > } > > + ioreq_request_mapcache_invalidate(p2m->domain); > + > return p2m_set_entry(p2m, gfn, INVALID_MFN, page_order, p2m_invalid, > p2m->default_access); > } > @@ -1301,6 +1304,8 @@ static int set_typed_p2m_entry(struct do > ASSERT(mfn_valid(mfn_add(omfn, i))); > set_gpfn_from_mfn(mfn_x(omfn) + i, INVALID_M2P_ENTRY); > } > + > + ioreq_request_mapcache_invalidate(d); > } > > P2M_DEBUG("set %d %lx %lx\n", gfn_p2mt, gfn_l, mfn_x(mfn)); > --- a/xen/arch/x86/mm/p2m-pod.c > +++ b/xen/arch/x86/mm/p2m-pod.c > @@ -20,6 +20,7 @@ > */ > > #include <xen/event.h> > +#include <xen/ioreq.h> > #include <xen/mm.h> > #include <xen/sched.h> > #include <xen/trace.h> > @@ -647,6 +648,8 @@ p2m_pod_decrease_reservation(struct doma > set_gpfn_from_mfn(mfn_x(mfn), INVALID_M2P_ENTRY); > p2m_pod_cache_add(p2m, page, cur_order); > > + ioreq_request_mapcache_invalidate(d); > + > steal_for_cache = ( p2m->pod.entry_count > p2m->pod.count ); > > ram -= n; > @@ -835,6 +838,8 @@ p2m_pod_zero_check_superpage(struct p2m_ > p2m_pod_cache_add(p2m, mfn_to_page(mfn0), PAGE_ORDER_2M); > p2m->pod.entry_count += SUPERPAGE_PAGES; > > + ioreq_request_mapcache_invalidate(d); > + > ret = SUPERPAGE_PAGES; > > out_reset: > @@ -997,6 +1002,8 @@ p2m_pod_zero_check(struct p2m_domain *p2 > /* Add to cache, and account for the new p2m PoD entry */ > p2m_pod_cache_add(p2m, mfn_to_page(mfns[i]), PAGE_ORDER_4K); > p2m->pod.entry_count++; > + > + ioreq_request_mapcache_invalidate(d); > } > } > > @@ -1315,6 +1322,8 @@ guest_physmap_mark_populate_on_demand(st > p2m->pod.entry_count -= pod_count; > BUG_ON(p2m->pod.entry_count < 0); > pod_unlock(p2m); > + > + ioreq_request_mapcache_invalidate(d); > } > > out: > --- a/xen/common/ioreq.c > +++ b/xen/common/ioreq.c > @@ -35,6 +35,17 @@ > #include <public/hvm/ioreq.h> > #include <public/hvm/params.h> > > +void ioreq_request_mapcache_invalidate(const struct domain *d) > +{ > + struct vcpu *v = current; > + > + if ( d == v->domain ) > + v->mapcache_invalidate = true; > + else if ( d->creation_finished ) > + for_each_vcpu ( d, v ) > + v->mapcache_invalidate = true; > +} > + > /* Ask ioemu mapcache to invalidate mappings. */ > void ioreq_signal_mapcache_invalidate(void) > { > @@ -206,6 +217,7 @@ bool vcpu_ioreq_handle_completion(struct > struct ioreq_server *s; > struct ioreq_vcpu *sv; > enum vio_completion completion; > + bool res = true; > > if ( has_vpci(d) && vpci_process_pending(v) ) > { > @@ -232,17 +244,27 @@ bool vcpu_ioreq_handle_completion(struct > break; > > case VIO_mmio_completion: > - return arch_ioreq_complete_mmio(); > + res = arch_ioreq_complete_mmio(); > + break; > > case VIO_pio_completion: > - return handle_pio(vio->req.addr, vio->req.size, > - vio->req.dir); > + res = handle_pio(vio->req.addr, vio->req.size, > + vio->req.dir); > + break; > > default: > - return arch_vcpu_ioreq_completion(completion); > + res = arch_vcpu_ioreq_completion(completion); > + break; > } > > - return true; > + if ( res && unlikely(v->mapcache_invalidate) ) > + { > + v->mapcache_invalidate = false; > + ioreq_signal_mapcache_invalidate(); > + res = false; > + } > + > + return res; > } > > static int ioreq_server_alloc_mfn(struct ioreq_server *s, bool buf) > --- a/xen/include/xen/ioreq.h > +++ b/xen/include/xen/ioreq.h > @@ -103,6 +103,7 @@ struct ioreq_server *ioreq_server_select > int ioreq_send(struct ioreq_server *s, ioreq_t *proto_p, > bool buffered); > unsigned int ioreq_broadcast(ioreq_t *p, bool buffered); > +void ioreq_request_mapcache_invalidate(const struct domain *d); > void ioreq_signal_mapcache_invalidate(void); > > void ioreq_domain_init(struct domain *d); > --- a/xen/include/xen/sched.h > +++ b/xen/include/xen/sched.h > @@ -225,6 +225,14 @@ struct vcpu > bool hcall_compat; > #endif > > +#ifdef CONFIG_IOREQ_SERVER > + /* > + * Indicates that mapcache invalidation request should be sent to > + * the device emulator. > + */ > + bool mapcache_invalidate; > +#endif > + > /* The CPU, if any, which is holding onto this VCPU's state. */ > #define VCPU_CPU_CLEAN (~0u) > unsigned int dirty_cpu; > @@ -444,11 +452,6 @@ struct domain > * unpaused for the first time by the systemcontroller. > */ > bool creation_finished; > - /* > - * Indicates that mapcache invalidation request should be sent to > - * the device emulator. > - */ > - bool mapcache_invalidate; > > /* Which guest this guest has privileges on */ > struct domain *target; >
On 17/02/2021 10:02, Jan Beulich wrote: > On 02.02.2021 16:14, Jan Beulich wrote: >> XENMEM_decrease_reservation isn't the only means by which pages can get >> removed from a guest, yet all removals ought to be signaled to qemu. Put >> setting of the flag into the central p2m_remove_page() underlying all >> respective hypercalls as well as a few similar places, mainly in PoD >> code. >> >> Additionally there's no point sending the request for the local domain >> when the domain acted upon is a different one. The latter domain's ioreq >> server mapcaches need invalidating. We assume that domain to be paused >> at the point the operation takes place, so sending the request in this >> case happens from the hvm_do_resume() path, which as one of its first >> steps calls handle_hvm_io_completion(). >> >> Even without the remote operation aspect a single domain-wide flag >> doesn't do: Guests may e.g. decrease-reservation on multiple vCPU-s in >> parallel. Each of them needs to issue an invalidation request in due >> course, in particular because exiting to guest context should not happen >> before the request was actually seen by (all) the emulator(s). >> >> Signed-off-by: Jan Beulich <jbeulich@suse.com> >> --- >> v2: Preemption related adjustment split off. Make flag per-vCPU. More >> places to set the flag. Also handle acting on a remote domain. >> Re-base. > > Can I get an Arm side ack (or otherwise) please? This looks good for me. Acked-by: Julien Grall <jgrall@amazon.com> Cheers, > > Thanks, Jan > >> --- a/xen/arch/arm/p2m.c >> +++ b/xen/arch/arm/p2m.c >> @@ -759,10 +759,9 @@ static void p2m_free_entry(struct p2m_do >> * has failed (error case). >> * So, at worst, the spurious mapcache invalidation might be sent. >> */ >> - if ( (p2m->domain == current->domain) && >> - domain_has_ioreq_server(p2m->domain) && >> - p2m_is_ram(entry.p2m.type) ) >> - p2m->domain->mapcache_invalidate = true; >> + if ( p2m_is_ram(entry.p2m.type) && >> + domain_has_ioreq_server(p2m->domain) ) >> + ioreq_request_mapcache_invalidate(p2m->domain); >> #endif >> >> p2m->stats.mappings[level]--; >> --- a/xen/arch/arm/traps.c >> +++ b/xen/arch/arm/traps.c >> @@ -1509,8 +1509,8 @@ static void do_trap_hypercall(struct cpu >> * Note that sending the invalidation request causes the vCPU to block >> * until all the IOREQ servers have acknowledged the invalidation. >> */ >> - if ( unlikely(curr->domain->mapcache_invalidate) && >> - test_and_clear_bool(curr->domain->mapcache_invalidate) ) >> + if ( unlikely(curr->mapcache_invalidate) && >> + test_and_clear_bool(curr->mapcache_invalidate) ) >> ioreq_signal_mapcache_invalidate(); >> #endif >> } >> --- a/xen/arch/x86/hvm/hypercall.c >> +++ b/xen/arch/x86/hvm/hypercall.c >> @@ -32,7 +32,6 @@ >> >> static long hvm_memory_op(int cmd, XEN_GUEST_HANDLE_PARAM(void) arg) >> { >> - const struct vcpu *curr = current; >> long rc; >> >> switch ( cmd & MEMOP_CMD_MASK ) >> @@ -42,14 +41,11 @@ static long hvm_memory_op(int cmd, XEN_G >> return -ENOSYS; >> } >> >> - if ( !curr->hcall_compat ) >> + if ( !current->hcall_compat ) >> rc = do_memory_op(cmd, arg); >> else >> rc = compat_memory_op(cmd, arg); >> >> - if ( (cmd & MEMOP_CMD_MASK) == XENMEM_decrease_reservation ) >> - curr->domain->mapcache_invalidate = true; >> - >> return rc; >> } >> >> @@ -327,9 +323,11 @@ int hvm_hypercall(struct cpu_user_regs * >> >> HVM_DBG_LOG(DBG_LEVEL_HCALL, "hcall%lu -> %lx", eax, regs->rax); >> >> - if ( unlikely(currd->mapcache_invalidate) && >> - test_and_clear_bool(currd->mapcache_invalidate) ) >> + if ( unlikely(curr->mapcache_invalidate) ) >> + { >> + curr->mapcache_invalidate = false; >> ioreq_signal_mapcache_invalidate(); >> + } >> >> return curr->hcall_preempted ? HVM_HCALL_preempted : HVM_HCALL_completed; >> } >> --- a/xen/arch/x86/mm/p2m.c >> +++ b/xen/arch/x86/mm/p2m.c >> @@ -28,6 +28,7 @@ >> #include <xen/vm_event.h> >> #include <xen/event.h> >> #include <xen/grant_table.h> >> +#include <xen/ioreq.h> >> #include <xen/param.h> >> #include <public/vm_event.h> >> #include <asm/domain.h> >> @@ -815,6 +816,8 @@ p2m_remove_page(struct p2m_domain *p2m, >> } >> } >> >> + ioreq_request_mapcache_invalidate(p2m->domain); >> + >> return p2m_set_entry(p2m, gfn, INVALID_MFN, page_order, p2m_invalid, >> p2m->default_access); >> } >> @@ -1301,6 +1304,8 @@ static int set_typed_p2m_entry(struct do >> ASSERT(mfn_valid(mfn_add(omfn, i))); >> set_gpfn_from_mfn(mfn_x(omfn) + i, INVALID_M2P_ENTRY); >> } >> + >> + ioreq_request_mapcache_invalidate(d); >> } >> >> P2M_DEBUG("set %d %lx %lx\n", gfn_p2mt, gfn_l, mfn_x(mfn)); >> --- a/xen/arch/x86/mm/p2m-pod.c >> +++ b/xen/arch/x86/mm/p2m-pod.c >> @@ -20,6 +20,7 @@ >> */ >> >> #include <xen/event.h> >> +#include <xen/ioreq.h> >> #include <xen/mm.h> >> #include <xen/sched.h> >> #include <xen/trace.h> >> @@ -647,6 +648,8 @@ p2m_pod_decrease_reservation(struct doma >> set_gpfn_from_mfn(mfn_x(mfn), INVALID_M2P_ENTRY); >> p2m_pod_cache_add(p2m, page, cur_order); >> >> + ioreq_request_mapcache_invalidate(d); >> + >> steal_for_cache = ( p2m->pod.entry_count > p2m->pod.count ); >> >> ram -= n; >> @@ -835,6 +838,8 @@ p2m_pod_zero_check_superpage(struct p2m_ >> p2m_pod_cache_add(p2m, mfn_to_page(mfn0), PAGE_ORDER_2M); >> p2m->pod.entry_count += SUPERPAGE_PAGES; >> >> + ioreq_request_mapcache_invalidate(d); >> + >> ret = SUPERPAGE_PAGES; >> >> out_reset: >> @@ -997,6 +1002,8 @@ p2m_pod_zero_check(struct p2m_domain *p2 >> /* Add to cache, and account for the new p2m PoD entry */ >> p2m_pod_cache_add(p2m, mfn_to_page(mfns[i]), PAGE_ORDER_4K); >> p2m->pod.entry_count++; >> + >> + ioreq_request_mapcache_invalidate(d); >> } >> } >> >> @@ -1315,6 +1322,8 @@ guest_physmap_mark_populate_on_demand(st >> p2m->pod.entry_count -= pod_count; >> BUG_ON(p2m->pod.entry_count < 0); >> pod_unlock(p2m); >> + >> + ioreq_request_mapcache_invalidate(d); >> } >> >> out: >> --- a/xen/common/ioreq.c >> +++ b/xen/common/ioreq.c >> @@ -35,6 +35,17 @@ >> #include <public/hvm/ioreq.h> >> #include <public/hvm/params.h> >> >> +void ioreq_request_mapcache_invalidate(const struct domain *d) >> +{ >> + struct vcpu *v = current; >> + >> + if ( d == v->domain ) >> + v->mapcache_invalidate = true; >> + else if ( d->creation_finished ) >> + for_each_vcpu ( d, v ) >> + v->mapcache_invalidate = true; >> +} >> + >> /* Ask ioemu mapcache to invalidate mappings. */ >> void ioreq_signal_mapcache_invalidate(void) >> { >> @@ -206,6 +217,7 @@ bool vcpu_ioreq_handle_completion(struct >> struct ioreq_server *s; >> struct ioreq_vcpu *sv; >> enum vio_completion completion; >> + bool res = true; >> >> if ( has_vpci(d) && vpci_process_pending(v) ) >> { >> @@ -232,17 +244,27 @@ bool vcpu_ioreq_handle_completion(struct >> break; >> >> case VIO_mmio_completion: >> - return arch_ioreq_complete_mmio(); >> + res = arch_ioreq_complete_mmio(); >> + break; >> >> case VIO_pio_completion: >> - return handle_pio(vio->req.addr, vio->req.size, >> - vio->req.dir); >> + res = handle_pio(vio->req.addr, vio->req.size, >> + vio->req.dir); >> + break; >> >> default: >> - return arch_vcpu_ioreq_completion(completion); >> + res = arch_vcpu_ioreq_completion(completion); >> + break; >> } >> >> - return true; >> + if ( res && unlikely(v->mapcache_invalidate) ) >> + { >> + v->mapcache_invalidate = false; >> + ioreq_signal_mapcache_invalidate(); >> + res = false; >> + } >> + >> + return res; >> } >> >> static int ioreq_server_alloc_mfn(struct ioreq_server *s, bool buf) >> --- a/xen/include/xen/ioreq.h >> +++ b/xen/include/xen/ioreq.h >> @@ -103,6 +103,7 @@ struct ioreq_server *ioreq_server_select >> int ioreq_send(struct ioreq_server *s, ioreq_t *proto_p, >> bool buffered); >> unsigned int ioreq_broadcast(ioreq_t *p, bool buffered); >> +void ioreq_request_mapcache_invalidate(const struct domain *d); >> void ioreq_signal_mapcache_invalidate(void); >> >> void ioreq_domain_init(struct domain *d); >> --- a/xen/include/xen/sched.h >> +++ b/xen/include/xen/sched.h >> @@ -225,6 +225,14 @@ struct vcpu >> bool hcall_compat; >> #endif >> >> +#ifdef CONFIG_IOREQ_SERVER >> + /* >> + * Indicates that mapcache invalidation request should be sent to >> + * the device emulator. >> + */ >> + bool mapcache_invalidate; >> +#endif >> + >> /* The CPU, if any, which is holding onto this VCPU's state. */ >> #define VCPU_CPU_CLEAN (~0u) >> unsigned int dirty_cpu; >> @@ -444,11 +452,6 @@ struct domain >> * unpaused for the first time by the systemcontroller. >> */ >> bool creation_finished; >> - /* >> - * Indicates that mapcache invalidation request should be sent to >> - * the device emulator. >> - */ >> - bool mapcache_invalidate; >> >> /* Which guest this guest has privileges on */ >> struct domain *target; >> >
--- a/xen/arch/arm/p2m.c +++ b/xen/arch/arm/p2m.c @@ -759,10 +759,9 @@ static void p2m_free_entry(struct p2m_do * has failed (error case). * So, at worst, the spurious mapcache invalidation might be sent. */ - if ( (p2m->domain == current->domain) && - domain_has_ioreq_server(p2m->domain) && - p2m_is_ram(entry.p2m.type) ) - p2m->domain->mapcache_invalidate = true; + if ( p2m_is_ram(entry.p2m.type) && + domain_has_ioreq_server(p2m->domain) ) + ioreq_request_mapcache_invalidate(p2m->domain); #endif p2m->stats.mappings[level]--; --- a/xen/arch/arm/traps.c +++ b/xen/arch/arm/traps.c @@ -1509,8 +1509,8 @@ static void do_trap_hypercall(struct cpu * Note that sending the invalidation request causes the vCPU to block * until all the IOREQ servers have acknowledged the invalidation. */ - if ( unlikely(curr->domain->mapcache_invalidate) && - test_and_clear_bool(curr->domain->mapcache_invalidate) ) + if ( unlikely(curr->mapcache_invalidate) && + test_and_clear_bool(curr->mapcache_invalidate) ) ioreq_signal_mapcache_invalidate(); #endif } --- a/xen/arch/x86/hvm/hypercall.c +++ b/xen/arch/x86/hvm/hypercall.c @@ -32,7 +32,6 @@ static long hvm_memory_op(int cmd, XEN_GUEST_HANDLE_PARAM(void) arg) { - const struct vcpu *curr = current; long rc; switch ( cmd & MEMOP_CMD_MASK ) @@ -42,14 +41,11 @@ static long hvm_memory_op(int cmd, XEN_G return -ENOSYS; } - if ( !curr->hcall_compat ) + if ( !current->hcall_compat ) rc = do_memory_op(cmd, arg); else rc = compat_memory_op(cmd, arg); - if ( (cmd & MEMOP_CMD_MASK) == XENMEM_decrease_reservation ) - curr->domain->mapcache_invalidate = true; - return rc; } @@ -327,9 +323,11 @@ int hvm_hypercall(struct cpu_user_regs * HVM_DBG_LOG(DBG_LEVEL_HCALL, "hcall%lu -> %lx", eax, regs->rax); - if ( unlikely(currd->mapcache_invalidate) && - test_and_clear_bool(currd->mapcache_invalidate) ) + if ( unlikely(curr->mapcache_invalidate) ) + { + curr->mapcache_invalidate = false; ioreq_signal_mapcache_invalidate(); + } return curr->hcall_preempted ? HVM_HCALL_preempted : HVM_HCALL_completed; } --- a/xen/arch/x86/mm/p2m.c +++ b/xen/arch/x86/mm/p2m.c @@ -28,6 +28,7 @@ #include <xen/vm_event.h> #include <xen/event.h> #include <xen/grant_table.h> +#include <xen/ioreq.h> #include <xen/param.h> #include <public/vm_event.h> #include <asm/domain.h> @@ -815,6 +816,8 @@ p2m_remove_page(struct p2m_domain *p2m, } } + ioreq_request_mapcache_invalidate(p2m->domain); + return p2m_set_entry(p2m, gfn, INVALID_MFN, page_order, p2m_invalid, p2m->default_access); } @@ -1301,6 +1304,8 @@ static int set_typed_p2m_entry(struct do ASSERT(mfn_valid(mfn_add(omfn, i))); set_gpfn_from_mfn(mfn_x(omfn) + i, INVALID_M2P_ENTRY); } + + ioreq_request_mapcache_invalidate(d); } P2M_DEBUG("set %d %lx %lx\n", gfn_p2mt, gfn_l, mfn_x(mfn)); --- a/xen/arch/x86/mm/p2m-pod.c +++ b/xen/arch/x86/mm/p2m-pod.c @@ -20,6 +20,7 @@ */ #include <xen/event.h> +#include <xen/ioreq.h> #include <xen/mm.h> #include <xen/sched.h> #include <xen/trace.h> @@ -647,6 +648,8 @@ p2m_pod_decrease_reservation(struct doma set_gpfn_from_mfn(mfn_x(mfn), INVALID_M2P_ENTRY); p2m_pod_cache_add(p2m, page, cur_order); + ioreq_request_mapcache_invalidate(d); + steal_for_cache = ( p2m->pod.entry_count > p2m->pod.count ); ram -= n; @@ -835,6 +838,8 @@ p2m_pod_zero_check_superpage(struct p2m_ p2m_pod_cache_add(p2m, mfn_to_page(mfn0), PAGE_ORDER_2M); p2m->pod.entry_count += SUPERPAGE_PAGES; + ioreq_request_mapcache_invalidate(d); + ret = SUPERPAGE_PAGES; out_reset: @@ -997,6 +1002,8 @@ p2m_pod_zero_check(struct p2m_domain *p2 /* Add to cache, and account for the new p2m PoD entry */ p2m_pod_cache_add(p2m, mfn_to_page(mfns[i]), PAGE_ORDER_4K); p2m->pod.entry_count++; + + ioreq_request_mapcache_invalidate(d); } } @@ -1315,6 +1322,8 @@ guest_physmap_mark_populate_on_demand(st p2m->pod.entry_count -= pod_count; BUG_ON(p2m->pod.entry_count < 0); pod_unlock(p2m); + + ioreq_request_mapcache_invalidate(d); } out: --- a/xen/common/ioreq.c +++ b/xen/common/ioreq.c @@ -35,6 +35,17 @@ #include <public/hvm/ioreq.h> #include <public/hvm/params.h> +void ioreq_request_mapcache_invalidate(const struct domain *d) +{ + struct vcpu *v = current; + + if ( d == v->domain ) + v->mapcache_invalidate = true; + else if ( d->creation_finished ) + for_each_vcpu ( d, v ) + v->mapcache_invalidate = true; +} + /* Ask ioemu mapcache to invalidate mappings. */ void ioreq_signal_mapcache_invalidate(void) { @@ -206,6 +217,7 @@ bool vcpu_ioreq_handle_completion(struct struct ioreq_server *s; struct ioreq_vcpu *sv; enum vio_completion completion; + bool res = true; if ( has_vpci(d) && vpci_process_pending(v) ) { @@ -232,17 +244,27 @@ bool vcpu_ioreq_handle_completion(struct break; case VIO_mmio_completion: - return arch_ioreq_complete_mmio(); + res = arch_ioreq_complete_mmio(); + break; case VIO_pio_completion: - return handle_pio(vio->req.addr, vio->req.size, - vio->req.dir); + res = handle_pio(vio->req.addr, vio->req.size, + vio->req.dir); + break; default: - return arch_vcpu_ioreq_completion(completion); + res = arch_vcpu_ioreq_completion(completion); + break; } - return true; + if ( res && unlikely(v->mapcache_invalidate) ) + { + v->mapcache_invalidate = false; + ioreq_signal_mapcache_invalidate(); + res = false; + } + + return res; } static int ioreq_server_alloc_mfn(struct ioreq_server *s, bool buf) --- a/xen/include/xen/ioreq.h +++ b/xen/include/xen/ioreq.h @@ -103,6 +103,7 @@ struct ioreq_server *ioreq_server_select int ioreq_send(struct ioreq_server *s, ioreq_t *proto_p, bool buffered); unsigned int ioreq_broadcast(ioreq_t *p, bool buffered); +void ioreq_request_mapcache_invalidate(const struct domain *d); void ioreq_signal_mapcache_invalidate(void); void ioreq_domain_init(struct domain *d); --- a/xen/include/xen/sched.h +++ b/xen/include/xen/sched.h @@ -225,6 +225,14 @@ struct vcpu bool hcall_compat; #endif +#ifdef CONFIG_IOREQ_SERVER + /* + * Indicates that mapcache invalidation request should be sent to + * the device emulator. + */ + bool mapcache_invalidate; +#endif + /* The CPU, if any, which is holding onto this VCPU's state. */ #define VCPU_CPU_CLEAN (~0u) unsigned int dirty_cpu; @@ -444,11 +452,6 @@ struct domain * unpaused for the first time by the systemcontroller. */ bool creation_finished; - /* - * Indicates that mapcache invalidation request should be sent to - * the device emulator. - */ - bool mapcache_invalidate; /* Which guest this guest has privileges on */ struct domain *target;
XENMEM_decrease_reservation isn't the only means by which pages can get removed from a guest, yet all removals ought to be signaled to qemu. Put setting of the flag into the central p2m_remove_page() underlying all respective hypercalls as well as a few similar places, mainly in PoD code. Additionally there's no point sending the request for the local domain when the domain acted upon is a different one. The latter domain's ioreq server mapcaches need invalidating. We assume that domain to be paused at the point the operation takes place, so sending the request in this case happens from the hvm_do_resume() path, which as one of its first steps calls handle_hvm_io_completion(). Even without the remote operation aspect a single domain-wide flag doesn't do: Guests may e.g. decrease-reservation on multiple vCPU-s in parallel. Each of them needs to issue an invalidation request in due course, in particular because exiting to guest context should not happen before the request was actually seen by (all) the emulator(s). Signed-off-by: Jan Beulich <jbeulich@suse.com> --- v2: Preemption related adjustment split off. Make flag per-vCPU. More places to set the flag. Also handle acting on a remote domain. Re-base. --- I'm still unconvinced of moving the flag setting into p2m_set_entry(). Besides likely causing false positives, we'd also need to make the function retrieve the prior entry's type in order to do this for displaced RAM entries only. Instead I've identified more places where the flag should be set.